In his Begriffsschrift (1879), Gottlob Frege presents both the first statement of the famous Frege’s Puzzle and the first, metalinguistic solution to it, based on the idea that the difference in cognitive significance between identity statements of the form “a=a” and “a=b” is due to difference in names. In Uber Sinn und Bedeutung (1892) he abandons this view in favor of a solution based on his distinction between sense and reference of the proper name. Despite many acknowledged problems of this second Fregean solution (see e.g. Dąmbska 1949, Kripke 1979, 1980, Kaplan 1989), his metalinguistic approach had been, until recently (Perry 2019, de Ponte et al. 2021), unjustly neglected. In our talk, we shall discuss the prospects of resurrecting Frege’s original solution using the framework of Kaplan’s (1990, 2011) metaphysics of words, following the work of Tarnowski and Głowacki (2022). At first, we shall develop Kaplan’s naturalistic idea of proper names as historical collections of uses essentially referring to the name’s bearer. In our talk, we discuss two main problems raised by Frege (1892): the problem of arbitrary link between the name and its bearer, and the problem of the possibility of a posteriori knowledge of true identity statements. We show that the two problems have plausible solutions within Kaplan’s framework. In the end, we discuss the modal properties of names in our theory, especially with respect to the hotly debated discussion between metaphysical stances of necessitism and contingentism (Williamson 2013).